Abstract: Based on the idea of Proper Orthogonal Decomposition (POD), Singular Value Decomposition (SVD) is used to decompose the solution space of physical problems into geometric subspace and design subspace, and random variable response results are obtained by linearly combining geometry and design subspace. Different from the traditional acceleration algorithm, the radial basis function (RBF) is adopted to approximate the design of the subspace response, the compressed expression of system information is realized, and the cost of calculation is reduced effectively. The boundary element method (BEM) is used for numerical analysis of two-dimensional potential problems, and the combination of SVD and RBF methods accelerates the rapid response analysis of BEM-based potential problems. Finally, a circular structure example is used to test the correctness and effectiveness of the algorithm developed.